[glatt]

So the investigators yesterday came out with a little more information.

They confirmed that this particular train was in automatic mode, and was being driven by the computer. They also confirmed that the emergency brakes had been applied. The button had been depressed in the driver's cabin, and the brakes themselves had the bluing that you would expect from emergency application. Also, the tracks had "skid marks" on them.

What they didn't say was how long the skid marks were, or how early the brakes had been applied.

It's frustrating that they didn't include this important information, so I tried to figure it out myself.

If you go to Google Earth and look at the accident scene, you can measure from the point of impact back in a straight line until that straight line gets obstructed by something, and then you know how far away the driver should have been able to see the stopped train. When you do this, you will see that the accident happened at a curve, under a bridge, and the visibility was actually pretty bad. The driver, if she had been paying 100% attention to the track in front of her, would have first seen the corner of the stopped train when she was about 355 meters away. At that particular location, according to the Washington Post, the train speed limit is 59 miles per hour. A train traveling at 59 miles per hour will cover 355 meters in 13.5 second. She wouldn't have seen the full train until she was about 160 meters away or 6 seconds from impact.

I don't know how long it takes to stop a train traveling at that speed. In normal operation, the trains take a while to stop, but they are trying to do it gently for the passengers' comfort.

So she had 13.5 seconds to see the stopped train, realize it was stopped, notice her own train wasn't stopping, still wasn't stopping, and slam on the emergency brake. She did all those things, but apparently not fast enough.

The head of Metro said that there is no evidence that she did anything wrong. But autopsy results, blood tests, cell phone, and texting records haven't been released yet.

The train had an unusual configuration. It had a lead car that normally isn't used as a lead car. In theory, it should have worked fine, but trains were seldom set up that way. And that lead car was overdue for brake maintenance. It's possible that the unusual configuration caused the autopilot to misread the situation and not stop.

I mentioned multiple layers of failure before. We now have: unusual train configuration, possible bad brakes, train stopped on a curve, and a "driver" who may have taken a couple seconds too long to apply the emergency brakes.